DeepAI AI Chat
Log In Sign Up

Improved Regularization and Robustness for Fine-tuning in Neural Networks

by   Dongyue Li, et al.
Northeastern University

A widely used algorithm for transfer learning is fine-tuning, where a pre-trained model is fine-tuned on a target task with a small amount of labeled data. When the capacity of the pre-trained model is much larger than the size of the target data set, fine-tuning is prone to overfitting and "memorizing" the training labels. Hence, an important question is to regularize fine-tuning and ensure its robustness to noise. To address this question, we begin by analyzing the generalization properties of fine-tuning. We present a PAC-Bayes generalization bound that depends on the distance traveled in each layer during fine-tuning and the noise stability of the fine-tuned model. We empirically measure these quantities. Based on the analysis, we propose regularized self-labeling – the interpolation between regularization and self-labeling methods, including (i) layer-wise regularization to constrain the distance traveled in each layer; (ii) self label-correction and label-reweighting to correct mislabeled data points (that the model is confident) and reweight less confident data points. We validate our approach on an extensive collection of image and text data sets using multiple pre-trained model architectures. Our approach improves baseline methods by 1.76 classification tasks and 0.75 target data set includes noisy labels, our approach outperforms baseline methods by 3.56


page 1

page 2

page 3

page 4


Robust Fine-Tuning of Deep Neural Networks with Hessian-based Generalization Guarantees

We consider transfer learning approaches that fine-tune a pretrained dee...

Fine-Tuning Pre-trained Language Model with Weak Supervision: A Contrastive-Regularized Self-Training Approach

Fine-tuned pre-trained language models (LMs) achieve enormous success in...

Self-Repetition in Abstractive Neural Summarizers

We provide a quantitative and qualitative analysis of self-repetition in...

CIFT: Crowd-Informed Fine-Tuning to Improve Machine Learning Ability

Item Response Theory (IRT) allows for measuring ability of Machine Learn...

Trainable Projected Gradient Method for Robust Fine-tuning

Recent studies on transfer learning have shown that selectively fine-tun...

Train and You'll Miss It: Interactive Model Iteration with Weak Supervision and Pre-Trained Embeddings

Our goal is to enable machine learning systems to be trained interactive...

Background Splitting: Finding Rare Classes in a Sea of Background

We focus on the real-world problem of training accurate deep models for ...